OGM2PGBM: Robust BIM-based 2D-LiDAR localization for lifelong indoor navigation

Abstract

Several studies rely on particle filter (PF) algorithms for robot localization in Occupancy Grid Maps (OGMs) extracted from building information models (BIM models). However, most ignore the possible discrepancies between the reference model and the real world (Scan-BIM deviations). These deviations affect the accuracy of PF drastically. This paper proposes an open-source method to generate appropriate Pose Graph-based maps from BIM models for robust 2D-LiDAR localization in changing and dynamic environments. First, 2D OGMs are generated from complex BIM models allowing autonomous navigation. Then, a technique converts these maps into Pose Graph-based maps enabling accurate pose tracking. Finally, a robust localization is proposed with a combination of state-of-the-art algorithms. We found that Pose Graph-based algorithms are four times more accurate than PF algorithms by tracking the robot's pose in a real environment. The proposed method contributes to a robust localization with a BIM model in changing and dynamic environments.